This article explains how writing for AI search is different from writing for SEO, and how it overlaps. It covers what has stayed the same and what you’ll need to tweak about how you write, so large language models (LLMs) can understand, summarise and reuse it accurately. If you want your content to appear in AI-generated answer boxes or summaries, this overview will give you the core principles you need and a to-do list of adjustments to make.
This article explains how writing for AI search is different from writing for SEO, and how it overlaps. It covers what has stayed the same and what you’ll need to tweak about how you write, so large language models (LLMs) can understand, summarise and reuse it accurately. If you want your content to appear in AI-generated answer boxes or summaries, this overview will give you the core principles you need and a to-do list of adjustments to make.
Hot off the press last month, a McKinsey poll found that half of consumers are already using AI‑powered search tools. And in 2028, it looks like those searches could lead to US$750 billion in annual revenue. That’s billion, with a B.
We can’t ignore it, of course. But the hype around AI reminds me very much of the early days of SEO. All at once, the marketing industry seemed to throw away everything we knew: that marketing is about understanding and connecting with people. Instead, we began focusing on understanding and connecting with Google. What we ended up with was article after terrible article written for no one and offering no value. Those articles ran out of SEO juice fast. As Google fine-tuned its algorithms to reward well-targeted, comprehensive, valuable content, those hacky, keyword-skip-bins plummeted in the ranks.
So, let’s agree to not do that again.
AI is great, but it doesn’t mean throwing out everything that came before. Let’s go over what’s changed, what’s stayed the same, and what – if anything – you need to do about it.
Far from it. SEO is evolving, and Google still handles around 90% of search queries around the world.
Writing for AI search is pretty similar to writing for SEO. If you’re already creating searchable, valuable content, a lot of it will be pretty good for AI search too. Here’s what to keep doing.
Google’s more recent algorithm updates have focused on user intent rather than literal keyword matches. Its systems try to understand what people actually want to know – not just the words they type.
So, Google will surface articles about Melania when I search for ‘Donald Trump wife’, even if those exact words don’t appear on the page.
User search: ‘CRM NZ under $50’ in…
Old-school Google
Keyword-matching: serves up a page with title, ‘CRM pricing models’.
An LLM
Intent focus: includes a page that specifically mentions ‘The best CRMs in New Zealand under $50 per month.’
That’s because, way back in 2015, Google introduced an AI-driven system called RankBrain, which started teaching its search engine to interpret meaning and context rather than pure keyword strings. RankBrain was kind of the first step towards AI-powered search. Modern large language models (LLMs) take that idea into hyperdrive. They figure out intent across multiple layers. Say you search “What’s the best CRM in NZ that integrates with Xero and costs under $50 a month?” The model will guess that you mean “content management system” and “New Zealand,” and then will look across content to come up with an answer that delivers on all your criteria.
That doesn’t change our job as writers, but it does intensify the requirements: get more targeted, understand readers more deeply and deliver value more fully.
The key out-take: Everything you already know about good content still applies. Think about user intent, clarity, depth and authority –they’re key for search engines, AI tools and (the important part) people.
Pearl-clutching articles tell us millennials and Gen Z don’t read. But long, information-dense articles perform better from a search perspective. And the reasons are obvious: long articles encourage dwell time, have more keywords and more information worth sharing. But people don’t want to read walls of text. Those dense articles need to be broken into digestible chunks. Again, this isn’t news, and hasn’t changed much with the arrival of AI search. Clear, descriptive subheads, logical flow, highlighted juicy stats and quotes – these make text more accessible for readers and machines.
The key out-take: Dense, valuable content wins, but only if it’s structured so both humans and machines can navigate it easily.
Yes, yes, thanks, Google, for the truly uninspired anagram: EEAT. It stands for Experience, Expertise, Authoritativeness and Trustworthiness – basically a poor repackaging of branding basics. Content does best (and get this) when it comes from a credible source. I know – revolutionary.
Sorry if I sound annoyed. It’s because I am.
Anyway, the AI algos do the same. They look for signs that what you’re putting out was written by a real person and based on real expertise.
How do you deliver on EEAT for SEO while nailing GEO? (Acronyms extreme!):
- Lean hard into your brand voice. Your writing needs to sound distinct and human (ie, not like Chat GPT).
- Reference primary sources and include original quotes from real people.
- Add a byline to every article – WFB didn’t write this, I did. Let people click through to a bio.
- Stay in your niche. It’s why you won’t catch me writing about marketing in general.
- Give writers access to your experts. Your copy needs to go beyond desktop research.
The authenticity markers that AI notices:
These signals tell models that your writing reflects real expertise, not generic filler:
- Direct quotes from real people
- Real-world examples and anecdotes
- Numbers like stats, dollar amounts and dates
- A byline linked to a real bio
In SEO, like in love, hygiene won’t be the reason you’re picked, but it can be the reason you aren’t.
That hasn’t changed. Good titles, descriptions, tags and schema make things easy for search engines and AI systems to understand your content and serve it up to the right people. Load times and clean code tell the machines you’re likely to deliver a good user experience. It’s all just stuff you need to keep on top of, like laundry.
Obviously, there are some things you need to tweak about the way you write to optimise for AI. Luckily, these things will also play nicely with SEO.
SEO’s shift into assessing user intent is amplified by AI, because it’s just much better at it. You still need keywords to show what you’re talking about, and those will happen in any decently focused bit of writing. AI can also look at the meaning behind phrases – it doesn’t matter so much how many times those keywords appear. That’s great news for writers. We don’t need to manage the tension of finding places for clumsy key phrases, and can switch back to our home base: natural language, synonyms and context to add meaning. It’s why I haven’t just repeated “writing for AI search” over and over. I’ve used ‘semantic variants’ like AI-search optimisation, LLM-optimisation, AI-friendly text and AI search ranking factors
I’d probably have done this anyway because it’s nicer to read, and (yay) is ideally suited to AI search.
With SEO, you want Google to serve up your whole page when people search with your chosen keywords. AI won’t do that. Instead, it will serve up a snippet, hopefully with your page linked as a reference. Current advice is that each paragraph should make sense on its own. My feeling is that as the tools get more sophisticated, this will become less important. The algos and LLMs want to serve up human, interesting and valuable content, not writing that’s been made less readable to cater to robot requirements.
So make paragraphs self-contained as long as you can do it without sacrificing readability and human voice.
AI tools seem to swing widely between spine-chilling brilliance and being just…really, really dumb. The answers you get served in ChatGPT may look logical and human, but they’re not. They need to be told how to read a document – what’s the main point of this text? How does everything fit together? That’s why schema markups are so important and why content that’s structured like a user manual works best. Subtly linked, richly metaphorical or allegorical prose might be a shoo-in for the Man Booker, but it’s too cerebral for the bots.
It looks as though AI search platforms prefer fresher content. One clever-clogs ran the numbers and found that over three‑quarters of ChatGPT’s top‑cited pages had been updated in the last 30 days, so upload and update regularly.
Ultimately, AI-optimised writing is just good writing with a few strategic twiddly bits. Write something valuable, original and targeted, then tweak to help LLMs read it and SEO algos find it. Luckily, those things work well for readers too: clear structure, direct answers, explicit facts, helpful headers and standalone paragraphs. My tip: get ChatGPT to tell you what tweaks you could make. That feels poetically appropriate. The LLM can advocate for robot needs, and leave the humans to me.
Now for the call to action: send us your writing brief to get it found in search, quoted by AI and read by actual humans.
So, here are my FAQs about writing for AI search. This is an easy way for you, the reader, to understand this article and jump straight to the good parts. It’s also an extremely convenient way of delivering exactly what AI search loves.
These FAQs summarise the main ideas in the article, in a format preferred by AI tools: conversational questions, with concise, self-contained answers.
The biggest difference between writing for SEO and AI search is this: SEO content needs to be machine-readable, AI content needs to be machine-reusable. The LLMs love lifting and shifting a paragraph to directly answer a user query, whereas SEO looks for signs that your content could answer that question. But SEO and AI search are more similar than they are different: great SEO writing works well for AI search, and vice versa.
Keywords still matter in AI search, but they’re used differently. In SEO, keywords directly influence what search terms to rank for. LLMs use keywords to give context clues – like, “What the hell is this blog about?” To write for this is functionally the same – write naturally around your topic using phrases people use when they ask questions.
Make your content show up in AI summaries by ensuring your site and your content are both credible and highly regarded online. That means you’re showing expertise backed by evidence, and you’re being linked to and mentioned by credible sources. Then, craft answers to questions clearly and completely in one short, self-contained paragraph. FAQ lists do this perfectly.
Absolutely, SEO tools like keyword planners are still super useful. They’re a great way to find out what people are asking about, so you create content that answers them. This isn’t all that dissimilar to writing for SEO – you find out what people are searching for and build content that delivers on that.
The writing that performs best in AI search is clear, human and helpful. It directly and concisely answers specific questions – the more targeted and granular, the better. Think about each section as a tiny self-contained blog that fully covers the topic, with no extras. Again, you should be doing this anyway if you’re smashing it at SEO.
Signal authority to AI systems the same way you do with humans: show your experience and include real info, not just opinions (think examples and data). Backlinks are still important, but LLMs also look for signs that you know what you’re talking about, that the content is original and that you’ve verified your claims.
Yes, long-form content is useful – and can even work best, but only if it’s laid out well for skim readers. Avoid the wall of text at all costs. Readers, Google and LLMs will all thank you for chunked-down content, descriptive subheads, FAQs and call-out sections.
Headings that are optimised for readers will also be optimised for AI. They act as sign posts, indicating what content is where on the page, so readers can skip to the most relevant bits, and AI understands what it should be serving up to searchers. AI likes headers that mimic searched questions, but it can make pretty clunky reading if each subhead is a question. In the long run, it’s always better to serve your reader – the LLMs will catch up.
AI search won’t make SEO obsolete – it’s the next evolution. The fundamentals haven’t changed: understand your audience, offer value, and write clearly. That’ll suit both the search algorithms and the LLMs, but for different reasons.
No, it doesn’t matter if the FAQs repeat content from the main article. Think about this as the TL:DR list – you’re doing the job of producing an AI summary before the bots get their hands on your article.
The best practices for AI-optimised writing are very similar to best-practice writing: targeted, authentic, credible, clear and useful. Once you’ve done that, you can work on the LLM to-do list – add a summary or intro section upfront that directly ‘answers’ the article or page headline, make each paragraph work in isolation, and use question headers where you can.
The lead copywriter and creative director at Words for Breakfast. She spends much of her time working with the copy writing team across a variety of projects.
